Windows 2008R2 HyperV Driver Crash BSOD

[JHenn6]

Hello.

1/ I have the same problem with 2 systems:

• config1quad opteron 8347HE64GB will be extended to 96GB2 Intel dual port ET 82576 E1G42ETmobo Tyan S4989

• config2quad opteron 835032GB will be extended to 64GB2 Intel dual port ET 82576mobo Tyan S4985-E

Windows 2008 R2 HyperV (SP1 do not solve problem)

need VMQ and VMBDrivers 14.7 OK.

Drivers 15.0, 15.5, 16.0 BSOD as soon as HyperV services are enabled and VMs start.

driver WinUpdate via MS crash!I have tried many different parameters tweaks. No luck as soon as VM Queues are enabled.

Any idea?

2/ My customer want to buy many quad ET 82576 E1G44ET2

and 82599 X520-T2 E10G42BT for actual systems

and new systems based on Tyan S8812, S8225,S8232, Supermicro H8QG6-F-O and H8DGI-F-O

Do you think pb will be the same with theses interfaces/mobo?

Thank a lot for help.

Best RegardsDaniel

[Mark_H_Intel]

If you disable VMQ on the Ethernet adapters do the crashes stop?

[idata]

Same/similar problem here.

We do a full uninstall and then reinstall the driver version we want to try. Driver e1q62x64.sys dated Dec 2009 seem to work with VMDq enabled after some trouble getting it installed. Drivers dated July or Sep 2010 won't work with VMDq enabled at all. Over time, either version seems to cause issues in some of the VMs (mostly the file server will slowly lose connectivity, files are left open, progressively slower connections until server becomes unreacheable - it can take a few hours, or weeks/months between a restart and the server exhibiting this behavior) - even if we disable VMDq, there is something wrong that eventually surfaces (not to mention that there seems to be a performance problem).

We first get a BSOD when we assign the team to the Hyper-V Switch. We then manually uncheck Microsoft Virtual Switch protocol from the NIC, and we are then able to associate the team NIC to the Hyper-V Switch and then we're usually able to get VMs to come online w/o causing the host to BSOD.

We have seen that Windows 2008 and 7 VMs can get corrupted by this problem. So if you're seeing this problem make sure you have a backup and/or use a fresh VM each time you try (otherwise the BSOD will happen even if you have already solved the problem).

We have tried installing Intel driver set 15.5, 15.6, 15.7 and 16.1 (none of which seemed to work). We have also tried with Dell provided drivers Intel_LAN_12.1.0_W2K3_8_64_A02_15.4.91.0 (which has an e1q62x64.sys dated 2009/12) and Intel_LAN_12.5.2_W2K3_8_64_A00_15.4.3.2 (which has an e1q62x64.sys dated 2010/09).

We're sporadically getting event 28 from VMSMP that read:

Port '...' was prevented from using MAC address '...' because it is pinned to port '...'.

Obviously, we don't have any duplicate MAC addresses. The number of event 28 messages was more frequent when using Intel_LAN_12.1.0_W2K3_8_64_A02_15.4.91.0 (Intel_LAN_12.5.2_W2K3_8_64_A00_15.4.3.2 seemed to help).

We have teamed all 4 ports in a single team, or in 2 teams with 2 ports each (and manullay assigning VMs to either team). Seems to make no difference in the BSOD behavior nor in the degraded connectivity.

We were hoping that SP1 would help, but given your experience, it seems it won't.

Server is a Dell T710 with dual Xeon X5650 and 2 dual port Intel-ET network cards and 24 GB RAM running Windows 2008 R2.

[Mark_H_Intel]

jdh@activanos and zootie,

The BSOD is likely caused by an issue in the base driver (e1q). The next update of the driver will contain the fix. I will contact you via private messages with a workaround.

Mark H

[idata]

Mark,

I am experiencing the exact same issue as the user above. Is there an issue with the base driver as you mentioned? Any idea when they'll release a revision to it?

Is this an issue with VLMB being enabled in tandem with VMq?

I've purchased a total of 4 Dual Port ETs to do failover clustering and need to get this resolved so I can go forward.

Thanks

[Mark_H_Intel]

@jt002

The only work around I know for the BSOD in a clustering environment is to disable VMQ. I doubt that changing the team type (mode) will help since the issue is in the NIC driver.

Mark H

[BSchi1]

Hi,

We have the same Problem with BSOD when NIC Teaming (VMLB) and *vmq is enabled.

Is there any solution up to now?

We use an INTEL Modular Server with the 5520VI Compute Modul Firmware 6.7 and the BIOS Setting "Maximize/minimum Memory below 4GB" set to max.

Microsoft Hyper-V Server with Failovercluster is installed.

Greeds Björn

[Mark_H_Intel]

The e1q driver update for this issue has not been released yet. The only workaround for now is to not enable VMQ.

Mark H

[BSchi1]

Thanks for the fast reply Mark,

Do you know, when this problem will be fixed?

Björn

[idata]

Hello Mark,

can you say when the problem will be fixed ?

[Mark_H_Intel]

@Thombo

Great timing. Maybe your psychic. The web pack with the bug fixes (version 16.4) just went live today.

http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=18725 http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=18725 will get you the files you need to update the drivers and software.

[idata]

Great work, seems to work. i installed the new driver in my Hyper-V Cluster, enabled VMQ, no Bluescreens :-)

Thank you

[BSchi1]

IT Works.

Tried with new Proset 16.4, enabled vmq and it Works.

Thanks